The rapid expansion of artificial intelligence is driving an equally rapid buildout of data centres. Yet the future of AI infrastructure will depend on more than chips and computing capacity. Data centres need dependable electricity, water, land, connectivity and support from the communities around them. A new World Economic Forum report, developed with Oliver Wyman, argues that these conditions must shape decisions from the earliest stages of a project.
The scale of the challenge is substantial. The report estimates that global data centre investment could reach $7 trillion by 2030, while electricity consumption could more than double between 2025 and 2030. Much of this demand will be concentrated in particular regions, placing pressure on local grids and other shared resources.
Efficiency depends on location
A data centre can perform well on a standard efficiency measure and still create problems beyond its walls. Water used to generate its electricity may add to its footprint even if its cooling system uses little water onsite. Cooling choices also involve trade-offs: evaporative cooling can reduce electricity use while increasing water consumption; air cooling generally uses less water but requires more power. The right choice therefore depends on local conditions.
The report identifies six recurring constraints on data centre development: electricity, water, cooling, community acceptance, land and regulatory complexity. They often overlap. A promising site may have available land but limited grid capacity. Another may have ample power but face water stress or community concerns about resource use. Assessing each facility in isolation can miss the combined pressure that multiple projects place on a region.
Make the consequential decisions early
The report’s central recommendation is to address sustainability and resilience during site selection, design, procurement and financing. By the time a facility begins operating, many choices about its resource use and ability to adapt have already been made.
Its playbook recommends testing projects against future scenarios, including hotter conditions, changes in water availability, higher-density AI hardware and evolving electricity needs. It also encourages designs that can accommodate upgrades, such as future liquid cooling, water reuse or changes to power systems. Alongside familiar measures of energy and water efficiency, the report proposes tracking lifecycle carbon and water impacts, local water stress, climate exposure, service reliability and community effects.
A responsibility shared across the AI ecosystem
Developers and operators make critical infrastructure choices, but they are not the only decision-makers. Investors can assess resource and climate risks before financing a project. Policymakers can plan for regional capacity and require early community engagement. Organizations purchasing AI and cloud services can ask providers for clearer information about environmental impacts, consider where workloads run and choose an appropriate level of computing power for each task.
Original commentary
JA Logic Labs Perspective
For JA Logic Labs, the report points to a broader question for AI governance: how can the benefits of AI grow without placing unsustainable demands on the systems and communities that support it? Answering it requires attention to physical infrastructure as well as to the technology running inside it.
Sources
- World Economic Forum — Data Centre Sustainability and Resilience: A Decision Playbook for AI Infrastructure (opens in a new tab)
Published
Reported findings, estimates, and recommendations are drawn from the sources above. Commentary in the “JA Logic Labs Perspective” section reflects JA Logic Labs' own views.